© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Julius Han · Gated SAE for Llama3-8B-Instruct
    5. Llama3-8B-IT
    7. Residual Stream
    9. 25-RES-JH
    11. 63375
    Prev
    Next
    INDEX
    Explanations

    instances of the word "refuse" and its variations, indicating a theme of rejection or opposition

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    Juliushanhanhan/llama-3-8b-it-res/blocks.25.hook_resid_post
    Features
    65,536
    Data Type
    float32
    Hook Name
    blocks.25.hook_resid_post
    Hook Layer
    25
    Architecture
    gated
    Context Size
    1,024
    Dataset
    Juliushanhanhan/openwebtext-1b-llama3-tokenized-cxt-1024
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    reich
    -0.16
      ↵  ↵
    -0.15
    la
    -0.15
    ago
    -0.15
    aze
    -0.15
    ERM
    -0.14
    å¤ĩ
    -0.14
    cha
    -0.14
    away
    -0.14
    erm
    -0.14
    POSITIVE LOGITS
     to
    0.38
    /ref
    0.20
    entially
    0.18
    issippi
    0.17
    to
    0.17
    	to
    0.16
     permission
    0.15
    resher
    0.15
    _to
    0.15
     temptation
    0.15
    Activations Density 0.017%

    No Known Activations